M01142
Does Swa-explanation and Explanatory Feedback Mitigate
Halo Effect in Auditor Professional Judgment?
ABSTRACT
Previous research found that halo effect in audit context decrease quality of professional
judgment auditor. We propose that mitigating halo effect is necessary to improve accurate and
professional judgment of the auditor. We hypothesize that (1) Auditors with complex data would
determine the risk of material misstatement lower than auditors with non-complex data, (2) risk
of material misstatement for account that is determined by an auditor with self-explanation
higher than risk of material misstatement for account that is determined by an auditor without
self-explanation, (3) risk of material misstatement for account that is determined by an auditor
with explanatory feedback is higher than risk of material misstatement for account that is
determined by an auditor without explanatory feedback. Data from laboratory experiment using
various levels of auditors in Surabaya City Indonesia. The results show that halo effect occur
when auditor with complex data determine inaccurate professional judgment. The finding
suggests that shows that explanatory feedback as a strategy of mitigating halo effect, however we
do not find that self-explanation as a mitigating halo effect.
JEL: M42
KEYWORDS: Halo effect, complex data, non-complex data, self-explanation,
explanatory feedback, professional judgment
1
Introduction
AICPA guidance (2008) stated that assessing of client expectation is important phase in
analytical procedure. Analytical procedure is one of planning phase in understanding of client
business and industry. Holistic perspective in strategic assessment increases auditor attention to
client business risk and helps to increase the risk of misstatement (Bell, Peecher and Solomon,
1992). Research in psychology found that holistic evaluative in evaluating performance
associated with a holistic perspective potential to bias that known as the “ halo effect ”(Murphy
et al. 1993).
In audit context, the halo effect evidence support by O’Donnel and Schultz (2005) and
Grammling et al. (2010). The halo effect occurs when auditor assesses business risk in first step
of planning and then assesses risk of material misstatement that conform the value of business
risk (O’Donnel and Schultz 2005). These findings suggest that the holistic perspective auditors
develop during strategic assessment influences subsequent judgment. The other audit situation,
the halo effect occurs when knowledge about material weakness unrelated to the compensating
control encourages auditor to require a higher level of precision for a compensating control
(Grammling et al. 2010). Previous research investigate the predictor halo effect is holistic
perspective in evaluating an object. In audit context, we propose that the predictor of halo effect
is data complexity. Hogarth and Einhorn (1992) stated that the determinant of ambiguity is data
complexity, refers to the amount of information contained in each piece of data. If individual
weigh the judgment based on latest piece of data, known as recency effect. But, if the judgment
based on earlier piece of data is a primacy effect. Grick (2008) stated that halo effect is related
with primacy effect. So, this research proposes that the ambiguity of data caused halo effect in
analytical procedure.
2
Previous research (O’Donnel and Schultz 2005; Grammling et al. 2010) investigates the halo
effect in audit context but the mitigating strategy consider to research. Grammling et al. (2010)
stated that future research must considering mitigating strategy halo effect. This research
addresses the research question: how to mitigate halo effect in audit context? Cooper (1981b)
stated that there were several approaches to mitigate halo effect, for example, statistic control,
familiarity increasing by ratings, setting rating group, grouping irrelevant category, and rating
after observation and review. From those various methods, the method that is suitable for audit
context is training that is knowledge acquisition method. Training for beginner auditors is meant
to obtain complex analytical skill (Earley 2001). This is also in accordance with general standard
of first auditing stating that audit training should be done by a person or more who have adequate
skill and technical training as an auditor.
Beginner auditors learn procedural skills through personal experience or observation on other
more experienced auditors (Earley 2001). Research result from Earley (2001) extended Bonner
and Walker (1994) by conducting effectiveness testing of two knowledge acquisition methods,
which are self-explanation and explanatory feedback.
Self-explanation is a knowledge
acquisition method by giving individual an opportunity to explain to themselves the
rationalization answer that is chosen by them on some cases given. Explanatory feedback is
knowledge acquisition method by using an example and using it to solve a new case.
We experimentally examine auditor judgment in various levels (junior, senior, manager,
supervisor and partner) about the risk of material misstatement on sales account in noncomplexity and complexity data. The result suggests that the complexity data impact to auditor
professional judgment and the appropriate strategy to mitigate halo effect is explanatory
feedback. The next sections present literature review and hypotheses, followed by a description
3
of the experimental design. The results are then discussed, followed by conclusion and the
implication of research.
Literature Review
Thorndike (1920) defined halo effect as a marked tendency to think of the person in general as
rather good or rather inferior and to color the judgments of the (person’s specific performance
attributes) by this general feeling. Individual who conducts consideration for decision making
often has beginning assessment (anchor) on information that is then adjusted when getting new
information, that is called adjustment and anchoring heuristic (Hogarth, 1987). Auditor when
tends to weigh on last information is called recency effect, and when tends to weigh on the first
information is called primacy effect. Theory explaining primacy effect and recency effect is
known as belief adjustment theory that is developed by Hogarth and Einhorn (1992) who
considered ordering effect to examine interaction between duty characterization and information
process strategy. Belief adjustment model considers direction, power, and type of information
that by Bayes’ Theorem before, explained only decision making based on order and pattern of
information presentation. Halo effect, according to Grcic (2008), related to primacy effect
because the impressive anchor assessment was used as the assessment on additional new
information.
Individual does not revise their belief on additional new information because the
impressive anchor still attaches in their memory.
The Relation of Complexity Data and Risk of Material Misstatement
CPA firm develops a methodology for auditor to obtain, evaluate, and record evidence during
analytical procedure (Hirst and Koonce 1996; Trompeter and Wright 2006).
Holistic
information can be obtained from client or audit team leader after conducting understanding on
client’s business and given to senior auditor and junior auditor as a field assignment executor. If
4
auditor accepts holistic information that generally describes client’s condition that is convincing,
then in giving assessment on risk level of misstated accounting in analytical testing stage, it tends
to be in accordance with holistic information and the partner’s assessment.
This condition is supported that professional consideration done by experienced auditor would be
better than the one done by less experienced auditor (Knapp and Knapp, 2001). Executor auditor
as information provider tends to seek information that is consistent with his/her consideration,
meanwhile, partner as reviewer would tend to seek evidence that is inconsistent (Libby and
Trotman 1993). This can be explained by cognitive model of justification process which is
manipulation to seek evidence that supports other decisions or considerations (Peecher 1996).
Auditor combines objective justification in mental representation as quickly as they get, that
decision taken needs support (Gibbin dan Newton 1994). This condition occurs because in audit
methodology, decision making process is gradually, so the previous decision may affect the next
decision.
Phillips (1999) found that auditor who evaluating evidence related to low (high) - risk account
would be less (more) sensitive on detailed evidence of aggressive financial reports. Thus,
executor auditor when facing general information and obtaining strong impression on client’s
condition would tend to determine consideration that is consistent with the general assessment
when facing detailed transaction assignment.
Client’s condition that convinces and obtains low business risk value from partner as well as
good assessment of internal control system would cause auditor to determine low misstate risk.
Holistic information has formed mental representation on auditor about client who initially is
assessed positive. The limitation of auditor as an individual would tend to conduct assessment of
the next client generally similar to the beginning assessment. By representative bias, it can be
5
explained that general information is used as a comparison on the similarity of general
assessment with detailed attribute of financial reports in analytical assignment. When faced on
detailed assessment of accounting, auditor would use general information that is easily absorbed
in memory.
The level of material misstatement is considered high if auditor faces a client with high-risk
business. The determination of business risk based on the understanding on holistic information
of client that includes various aspects both from internal and external. Auditor facing holistic
information that is convincing on a client would tend to have high halo bias in determining the
risk of material misstatement in analytical testing. Client that is considered well by partner
would make mental representation in auditor that other evidences would in accordance with the
good beginning assessment. Auditor who conducts strategic assessment at high (low) level
would tend to determine the risk of material misstatement to be high (low) (O’Donnel dan
Schultz, 2005). Positive holistic on client’s condition in the form of high strategic assessment
could potentially cause high halo effect when facing account fluctuation inconsistency.
Hogarth and Einhorn (1992) introduced belief adjustment model that refers to the amount of
information contained in each piece of evidence. A greater number of the datum more
ambiguous unless a deeper level of cognitive processing eliminates the inaccurate hypotheses
(Luippold and Kida 2012). In managerial context, Booker et al. (2007) found that compared to
relative cost information increases designers’ focus on cost minimization. Specific cost
information were given the exact dollar cost of each design option. Dilla and Stone (1997)
suggest that individuals require more cognitive effort to acquire and use that are presented in
words.
6
Specific information that shows client’s condition in detail that there is the emergence of
fluctuation accounting inconsistence could potentially cause low halo effect.
Specific
information is pooled of data and can be presented in complex and non-complex data. Auditor
would be better in grasping the signal of account fluctuation inconsistency in specific
information (non-complex data). Therefore, implication of complex data in analytical assignment
which is the determination of high level risk of material misstatement. Based on the previous
argumentation and research, then, hypothesis one can be proposed as the following:
H1: Auditors with complex data would determine the risk of material misstatement lower than
auditors with non-complex data
The Relation of Self-explanation, Halo Effect, and Determination Risk of Material
Misstatement
Halo effect occurs when an auditor has an impression after facing holistic information, so it
affects mental representation in unconsciousness level. The effort of judgment bias mitigation is
done by emphasizing on alternative structure in evaluating information in decision making. The
effort is done because by general structure, it does not succeed (Balzer and Slusky 1992).
Research of consideration and decision making states that halo effect would affect sensitivity on
evidence that is contrary or not in accordance with holistic impression. Halo effect causes
individual to change the criteria used for evaluating and weighing decision information (Murphy
and Anhalt 1992). O’Donnel and Schultz (2005) gave empirical evidence in their experiment
when auditor conducting strategic assessment, then having halo effect, so they became
insensitive when facing account fluctuation that inconsistent in analytic testing.
Cooper (1981b) proposed training as one of methods to reduce halo effect. In accordance with
general standard of audit, to increase its professional consideration, CPA firm and Indonesian
7
Accountant Public Institute could hold training activities with various topics that in accordance
with audit assignment. Audit research shows that training could increase performance of audit
assignment including ratio analysis and real estate valuation (Bonner and Walker 1994; Earley
2001; 2003). Training context that can be used for auditors as the means of halo effect mitigation
is in analytical procedure because in this stage, halo effect could potentially emerges.
Determination of risk of material misstatement is done when an auditor conducts analytical
testing in planning stage. Auditors obtain information from both partner and convincing client,
so halo effect could potentially emerge. This bias affects auditor’s assessment when facing audit
evidence in detail. Therefore, the assignment of analytical procedure could be the focus of
training for halo effect mitigation.
Auditors who face complex data information have high potential of halo effect, so they tend to
be insensitive in determining the risk of material misstate.
Learning is a construction process
that makes a learner converses words and examples of instruction or that is presented in a text,
into skills that can be used to a problem solving (Chi et.al.1994). Learning is an acquisition and
knowledge organization through working memory that is a cognitive system to obtain and
manipulate information, intersection between long-term memory and information received
through senses, so it is able to be a significant component in the intellectual ability and the ability
problem solving (Brewster, 2011). By self-explanation knowledge acquisition method, auditors
with high halo effect could describe argumentation of risk level of misstatement given
previously. This description of process is a problem solving process, so the cognitive system
helps auditors to make a better professional consideration.
Based on argumentation and results of previous research, the third research hypothesis can be
proposed as the following:
8
H2: Risk of material misstatement with self-explanation is higher than risk of material
misstatement without self-explanation
The Relation of Explanatory Feedback Knowledge Acquisition Method, Halo Effect, and
Risk of Material Misstatement
Halo effect emerges because of holistic information received affects the professional
consideration when facing evidence in detail. Grammling et.al. (2010) emphasized the needs of
halo effect mitigation effort, so the quality of auditor professional consideration increased.
Kennedy (1993) explained the framework of unfamiliarity by using the method of training and
memory refreshing (internal data mitigation) and giving supporting instrument of decision
external data mitigation). Training by giving corrective feedback may mitigate halo effect
(Cooper 1981).
Balzer and Slusky (1992) showed that useful instrument to increase accuracy of decision making
is feedback. Lowe and Reckers (2000), Narsa (2010) used supporting instrument of foresight
decision to change auditor’s foresight perspective in order to approach hindsight perspective.
This is an effort to mitigate hindsight and foresight bias. In the effort of halo effect mitigation,
the use of decision supporting instrument that can be used is review result of audit manager on a
case faced by client.
Most of studies in psychology about learning were originally from working examples similar to
problems in the form of sentences on the textbook. Those examples are well-structured and each
had right answer objectively and solution procedure that can be accepted that was modeled in the
example.
Bonner and Walker (1994) showed that explanatory feedback and regulation
understanding showed that the result of cognitive psychology from learning that was wellstructured was also applied on audit domain that is ill structured. By explaining to auditors on
9
how to provide a right answer on a case helps them to obtain procedural knowledge. Training
may give good explanation before the problem, for example before working (similar to a case
given at the beginning of instruction) or after working (similar to a case after explanation
feedback).
Earley (2001) gave empirical evidence that an auditor who is given explanatory feedback would
have a better professional consideration compared to the one who does not get explanatory
feedback. Therefore, halo effect affects professional consideration, so when knowledge as the
form of explanatory feedback training method would increase professional consideration, as the
result, halo effect is mitigated.
Halo effect bias that is from individual limitation in processing information can be explained by
using heuristics representation. Individual tends to rely on information that has similarity level
with other information. To handle the bias, Bazerman (1994) explained the need of melting
process that changed status quo of individual decision making process in the form of emergence.
In explanatory feedback knowledge acquisition method, individuals would be faced to additional
information that can melt strong perception of memory attaching on the memory. Thus, halo
effect in risk determination of material misstatement can be reduced. Based on the argumentation
above, hypothesis 3 can be proposed as the following:
H3: Risk of material misstatement with explanatory feedback is higher than risk of material
misstatement without explanatory feedback.
10
METHODOLOGY
Experimental Design
The experimental design is pretest-posttest control group design. Shadish et al. (2002) explained
that the advantage of group control design with pretest and posttest is to prevent threat on
internal validity. An experiment is done by holding audit simulation in audit seminar. The
experiment in general can be described in research experiment matrix in table 1. The first factor
is data complexity consisting of two levels, non-complex and complex. The second factor is
method of knowledge acquisition that consists of self-explanation, explanatory feedback and
without self-explanation- explanatory feedback as a control group.
Experimental Subject
Experimental subject are auditors (junior, senior, manager and partner) in Surabaya City, the
second big city with large quantity of CPA firm in Indonesia. Randomization design is by
completely randomized design that by dividing two types of audit simulation modules to subjects
randomly. Therefore, each subject gets the same opportunity to be experimental group or control
group.
Operational Definition of Variable and Measurement
The complexity data has two levels, non-complex and complex data. The non-complex data was
adapted from Dilla and Stone (1997) and Booker, Drake and Heitger (2007). Participants with
non-complex data were presented the number of sales account information and words only,
without mentioning relative comparison. The complex data were presented in sales account
information in the form of number, words and relative comparison as Booker, Drake and Heitger
(2007).The words that explain the client situation in complex data contain explanation with
11
uncertainty condition. For example, the complex data stated that the increase of sales at 2012 is
17.76% relative with past year. The non-complex data stated that account of sales in 2012 has
increased. The risk of material misstatement on sales account in analytical test was measured by
the scores of 1 (very low) to 7 (very high). Participants determine material risk assessment in
three condition: (1) before get client data (risk of material misstatement on sales 1); (2) after get
client data (risk of material misstatement on sales 2), (3) after get mitigating strategy (risk of
material misstatement on sales 3).
RESULT AND DISCUSSION
Experiment
Sixty-five participants completed the study1. They were mainly staff-level auditor and partner
employed in Surabaya public accounting firms. The participants required to perform analytical
procedures on financial information with a seeded misstatement account. The participant got
simulation module consisting of module one for halo effect testing and complexity data and
module 2 for halo effect mitigation testing. The experimental design included two cases. Each
module was entitled and covered the same and had the same number of pages. The participants
consists of 38 junior auditor, 16 senior auditor, 3 manager, 6 supervisor and 2 partner.
Test of Demographic Characteristics on Risk of Material Misstatement on Sales Account 2
The risk of material misstatement of sales account 2 is determined after subject receiving
manipulation of holistic or non-complex data. ANOVA testing address to determine that risk of
material misstatement on sales account 2 is not affected by the difference of subject demographic
characteristic. The testing result by dependent variable was risk of material misstatement risk on
1
The participants invited to International Standard Auditing in Surabaya held on 2nd February 2013.
12
sales account 2 and independent variable that was demographic characteristic including sex, age,
position, working period, education, participation in public accountant training and audit training
show at table 2.
The result of ANOVA testing on the effect of subject demographic characteristic difference of
misstatement risk on sales account 2 showed that sex did not affect the risk of material
misstatement on sales account 2 (p=0.697). Age and position in CPA firm did not affect the risk
of material misstatement on sales account 2, and it was shown by probability value of age
(0.367) and position (0.886). Working period did not affect the risk misstatement on sales
account 2 that was shown by its probability value of 0.803. The last education, participation in
public accountant training and audit training also did not affect the risk misstatement on sales
account 2. The probability value of last education was 0.430, while public accountant training
was 0.938 and audit training was 0.773. Therefore, it could be concluded that risk of material
misstatement on sales account was not affected by the difference of subject demographic
characteristic. In experimental test, we must be careful with experimental error, i.e. accounting
and auditing capability of subjects. We test that difference of score of accounting and auditing
proficiency not affects risk of material misstatement on sales account 2. Testing of experimental
error (score of accounting and auditing proficiency) was done with ANCOVA (analysis of
covariate) with dependent variable is risk of material misstatement on sales account 2,
independent variable is scope of information and covariate variable is score of accounting and
auditing proficiency (Table 3).
13
Table 3 presents that score of accounting and auditing proficiency is not significant (0.630) and
scope of information as a manipulate variable is significant (0.014). The conclusion of this test is
risk of material misstatement on sales account 2 is only affected by scope of information and
does not affect the different score of accounting and auditing proficiency or demographic
characteristics.
Testing of Hypothesis 1
The result of hypothesis 1 testing is shown at table 4. The first hypothesis testing was done by
comparing the average of sales account 2 risk of material misstatement in the group with noncomplex data and the average risk of material misstatement on sales account 2 in the group with
complex data. Score pretest in cell with non-complex data is 4.875 and score posttest 4.875. The
distribution of non-complex data caused the increasing of sales account material misstatement
risk of 0.125. The opposite condition occurred in the group with complex data, with score
pretest 4.818 and 4.030, show the different of posttest-pretest the risk of material misstatement of
sales account -0.788. We compare the difference of risk of material misstatement on sales
account 2 and risk of material misstatement on sales account 1 between cell with complex data
and cell with non-complex data. The independent t-test shows that significance of that difference
is 0.049. This showed that subject with high halo effect and was given complex data, tend to
conform initial assessment.
The testing result showed that hypothesis 1 was supported by empirical evidence in this research.
Auditors who obtained non-complex data about the client would determine a lower risk of
material misstatement compared to the one who obtained non-specific data.
This finding
supported the occurrence of halo effect in the auditor receiving partner information about the
condition of client with complex or non-complex data. Higher halo effect and the complex data
14
from partner was still impressed by client’s well initial assessment, so the attached halo effect
would affect the risk assessment of material misstatement of sales account 2 to be similar to the
assessment of sales account 1 material misstatement risk. The opposite condition occurred on
auditor with high halo effect and obtaining information from partner with non-complex data.
Although it had positive impression in the initial assessment of the client, it would determine the
risk of material misstatement of sales account 2 higher than the risk of sales account 1 material
misstatement.
Testing Hypotesis 2
Hypothesis 2 Testing
Hypothesis 2 stated that self-explanation is a mitigation strategy to reduce halo effect. The
testing for pretest-posttest research design with control group was done by stages as the
following, first, determining the average difference of material misstatement risk of sales account
3 (posttest) with the average of material misstatement risk of sales account 2 (pretest), second,
comparing the difference of posttest-pretest in the group of non-complex data with the difference
of posttest-pretest in the group of complex data by independent t-test. The testing result of
independent t-test can be seen on table 5.
The testing result of independent t-test on the group of complex data showed that F value in
Levene Test was 0.096 with the significance value of 0.760 meaning that variances of both
groups were the same, while t-value was 0.073 with the significance probability of 0.942. This
finding showed that the average of material misstatement risk of sales account in cell 1 was not
significantly different with the average of material misstatement of cell 3 as control group.
15
The testing in the holistic group comparing the average difference of material misstatement risk
of sales account 3 to the average of material misstatement risk sales account 2 can be seen on
table 7. The testing result showed that F value in Levene test was 0.762 with the significance
value of 0.393 meaning that variances of both groups were the same, and t value was 0.660 with
significance probability of 0.516. This finding showed that there was no difference of material
misstatement risk in the group that was given self-explanation mitigation strategy with control
group that did not receive mitigation strategy. Therefore, it could be concluded that hypothesis 2
was not supported and self-explanation knowledge acquisition method did not get empirical
support as halo effect mitigation strategy.
Hypothesis 2 Discussion
Hypothesis 2 stated that self-explanation knowledge acquisition method would reduce halo effect
in determining material misstatement risk was not supported by empirical data. It means, selfexplanation did not affect significantly on the determination of accurate material misstatement
risk when auditor faced complex data that triggered high halo effect. The result of this research
did not succeed proposing self-explanation knowledge acquisition method as a mitigation
strategy to reduce the effect of halo effect in the determination of material misstatement.
The essence of self-explanation was to give auditor opportunity to realize the consideration that
was done and able to absorb information given by partner with complex data. Furthermore,
auditor was able to determine the risk of material misstatement more accurately compared to the
misstatement risk determined previously.
In this research, auditor with high halo effect was expected to be able to conduct self-explanation
and to revise professional consideration that had been given. This research gave the fact that
auditor who conducted self-explanation did not succeed to revise the professional consideration
16
that had been done. This was probably because the auditor was not able to realize their own
problem without other’s help.
Other findings in the group with self-explanation mitigation strategy was that it did not give right
rational argumentation with the case, in fact, some people answered only with one sentence
without trying to argue logically. The list of questions consisting 8 questions to memorize
condition of the client delivered in holistic or non-complex data of partner’s information given
before subject gave an argumentation. All questions were answered by the subject with the
average answer of 6.6, meaning that subjects were able to memorize well the information given.
However, some subjects when asked to give an argumentation, answered with only one or two
sentences.
Hypothesis 3 Testing
The existence of explanatory feedback knowledge acquisition as the mitigation strategy to reduce
the effect of halo effect in determining material misstatement risk was detected by determining
the difference of material misstatement risk average of sales account 3 (posttest) and the average
of material misstatement sales account 2 (pretest), and next, comparing the difference of posttestpretest risk average of non-complex data group. Theoretically, the group with complex data or
non-complex data that had high halo effect would change its assessment when receiving
explanatory feedback compared to the control group that did not receive explanatory feedback.
Hypothesis 3 testing was done by comparing the difference of material misstatement risk average
of sales account 3 with the average of material misstatement risk of sales account 2 for each
information scope group with independent t-test.
17
The testing showed that there was the significant difference between cell 2 that received
information scope and explanatory feedback mitigation strategy and cell 3 that received noncomplex data and without mitigation strategy. Statistically, it could be seen that the value of
t=2.422 with significance probability of 0.025 for the group with non-complex data.
The result testing in the group of complex data and receiving explanatory feedback mitigation
strategy showed that there was a significant difference between cell 5 (explanatory feedback) and
cell 6 (without mitigation). The result of t-test showed that t value=2.200 with p=0.04. The
testing showed that explanatory feedback knowledge acquisition method could be used as a halo
effect mitigation strategy, thus, the professional consideration taken was more accurate.
Hypothesis 3 Discussion
Explanatory feedback knowledge acquisition method would reduce halo effect in determining
material misstatement as hypothesis 3 was supported by empirical data. This finding was in line
with the previous research that explanatory feedback could increase the accuracy of decision
making (Balzer and Slusky, 1992; Earley, 2001; Bonner and Walker, 1994).
Halo effect occurred when the auditor had impression on the client’s condition that was
convincing and obtained partner’s information with complex data affecting mental representation
in unconsciousness level. The effect of halo effect occurring was the decreasing of sensitivity on
the evidence that was contrary or did not fit the first impression. In this research, the auditor was
asked to understand the business and industry of the client by watching a video and reading a
booklet of client’s company profile. The detection of halo effect was done by giving questions
to the auditor to give an initial assessment on minimarket management, distribution system and
financial performance of the client before receiving financial and non-financial information in
detail. The finding shows the majority of auditor had high halo effect.
18
The next stage was to ask the auditor understanding information from the partner about sales and
selling cost account and to conduct analytical procedure that compared quantitative and
qualitative information as well as compared the account balance in 2011 and 2012.
The
identification of the halo effect attachment when obtaining partner’s information so it affected
the accuracy of decision making could be seen from the material misstatement risk that was
determined in the group that received complex data with the risk of material misstatement on the
group that received non-complex data.
This research showed that the risk of material
misstatement determined in the group of complex data was lower than the risk of material
misstatement in the group of non-complex data.
The result of good initial assessment on the client was used as the basic to make decision when
conducting the next assessment (explaining representativeness heuristic). This could be reduced
by giving explanatory feedback in the form of manager’s review result containing the case
preview of client especially on the account fluctuation of sales and selling cost. The finding of
this result that was explanatory feedback could increase the accuracy of professional
consideration, so it reduced halo effect, strengthened the statement of Cooper (1981) that training
with corrective feedback could mitigate halo effect.
Conclusion
Initial judgment in evaluate convincing client profile will influence the subsequent judgment.
The result of this research found that halo effect occur in analytical procedure when auditor
determine professional judgment with complex data. Auditor with non-complex data will assess
risk of material misstatement lower than auditor with complex data.
19
Self-explanation knowledge acquisition method as halo effect mitigation strategy could not be
supported in this research. This finding showed that method of self-explanation with written
argumentation cannot change the first opinion and halo effect cannot mitigation. Cognitive
process from auditor’s self when conducting learning on a problem, could not be separated from
self-desire to learn and the willingness to analyze the problem.
The result of this research gave empirical evidence that explanatory feedback in the form of
manager’s review result was a knowledge acquisition method that could be halo effect mitigation
strategy on auditor. Halo effect that could be mitigated was marked by the increasing of
professional consideration accuracy in determining material misstatement risk. This finding was
consistent with (2001, 2003) that explanatory feedback could increase auditor’s professional
consideration.
This result also gave contribution that halo effect could be mitigated by
explanatory feedback.
This study has three limitations. First, about the experimental case materials just use positive
halo effect. Further research can extend with negative halo effect and determine risk of material
misstatement. Second, the context of analytical procedure is in initial hypothesis. Further
research is expected to improve the experiment materials in another audit test, i.e. in substantive
test or in analytical procedure in final evaluation phase.
Third, this research focus in an
individual judgment, other research may be developed in group decision making.
References
American Institute of Certified Public Accountants (AICPA). 2012. “AICPA Professional
Standards.” New York: AICPA
Bazerman, M.H. 1994. Judgment in Managerial Decision Making. Singapore: John Wiley and
Sons.
Balzer, W. and M. Sulsky. 1992. “Halo and Performance Appraisal Research: A Critical
Examination, “ Journal of Applied Psychology, p. 975-985
20
Booker, D. M., A. R. Drake, and D. L. Heitger. 2007. New Product Development: How Cost
Information Precision Affects Designer Focus and Behavior in a Multiple Objective
Setting. Behavioral Research in Accounting. 19: 19-41
Bonner, S. E., & P. L.Walker. 1994. The Effects of Instruction and Experience on the
Acquisition of Auditing Knowledge. The Accounting Review. 1: 157-178
Brewster, B. E. 2011. How a System Perspective Improves Knowledge Acquisition and
Performance in Analytical Procedures. The Accounting Review. 86 (3): 915-943
Cooper, W. H. 1981. “Ubiquitous Halo,”Psychological Bulletin.vol.90. no.2: p. 218-244
Chi, M.T.H., N. de Leeuw, M.H. Chiu, and C. Lavancher. 1994. Eliciting Self- Explanation
Improves Understanding. Cognitive Science. 18: 439-477
Dilla, W. N., and D.N.Stone. 1997. “Representations as Decision Aids: The Asymmetric Effects
of Words and Numbers on Auditors’ Inherent Risk Judgments,”Decision Sciences 28
(3): p.709-743
Earley, C. E. 2001. Expertise Acquisition in Auditing: Training Novice Auditors to Recognize
Cue Relationships in Real Estate Valuation. The Accounting Review.1: 81-97
______, C. E. 2003. A Note on Self Explanation as a Training Tools for Novice Auditors: The
Effects of Outcome Feedback Timing and Level of Reasoning on Performance.
Behavioral Research in Accounting. 15: 111-124
Gibbins, M, and J.D Newton. 1994. “An Empirical Exploration of Complex Accountability in
Public Accounting,”Journal of Accounting Research. (Autumn): p.165-186
Grammling, A., O'Donnel, E., and S.Vandervalde. 2010. “Audit Partner Evaluation of
Compensating Controls: A Focus on Design Effectiveness and Extent of Auditor
Testing,”Behavioral Research in Accounting. Vol. 29: p. 175-187
Hogarth, R.M. 1987. Judgment and Choice.2nd Edition. Singapore: John Wiley & Sons
Hogarth, R.M., and H.J. Einhorn. 1992. “Order Effects in Belief Updating: The BeliefAdjustment Model,” Cognitive Psychology, Vol. 24: p.278-288
Hirst, D., and L. Koonce. 1996. “Audit Analytical Procedures: A Field Investigation,”
Contemporary Accounting Research: 457-486
Libby, R and K.T.Trotman. 1993. “Audit Review as a Control for Biased Recall of Evidence in
Decision Making,”Accounting, Organizations and Society.18: p.559-574
Kennedy, J. 1993. Debiasing Audit Judgment with Accountability: A Framework and
Experimental Result. Journal of Accounting Research. 31: 481-491
Knapp, C., and Knapp, M. 2001. “The Effects of Experience and Explicit Fraud Risk Assesment
in Detecting Fraud with Analytical Procedures,” Accounting, Organization and
Society.vol.26: p.25-37
Luippold, B. L., and T.E. Kida. 2012. “The Impact of Initial Information Ambiguity on the
Accuracy of Analytical Review Judgment,” Auditing: A Journal of Practice & Theory. 31
(2): p.113-129
21
Lowe, D. J., and P. M. J. Reckers. 2000. The Use of Foresight Decision Aids in Auditor’s
Judgment. Behavioral Research in Accounting. 12: 97-118
Murphy, K., Jako, R.and R.L. Anhalt, 1993. “Nature and Consequences of Halo Error: A Critical
Analysis. Journal of Applied Psychology,” vol 2: p. 218-225
Murphy, K.R., and R.L Anhalt. 1992. Is Halo Error a Property of the Rater, the Ratees, or the
Specific Behavior Observed? Journal of Applied Psychology. 77: 494-500
Narsa, I, M. 2010. Alat Bantu Keputusan Foresight dan Strategi Mitigasi Bias Hindsight untuk
Mempersempit Kesenjangan di antara Perspektif Auditor dan Evaluator. Disertasi.
Universitas Gadjah Mada, Yogyakarta
O' Donnel, E. and Schultz, J. J. 2005. “The Halo Effect in Business Risk Audits: Can Strategic
Risk Assesment Bias Auditor Judment about Accounting Details?” The Accounting
Review, Vol 80, No 3: p. 921-939
Peecher, M.E. 1996. “The Influence of Auditor’s Justification Processes on Their Decisions: A
Cognitive Model and Experimental Evidence,”Journal of Accounting Research. 34(1):
p.125 – 140
Philips, F. 1999. “Auditor Attention to and Judgment of Aggressive Financial Reporting,”
Journal of Accounting Research. Vol 37. No.1. Spring
Shadish, R.W; Cook, T and Campbell, D.T. 2002. Experimental and Quasi-Experimental Design
for Generalized Causal Inference. Houghton Mifflin Company
Thorndike, E. L. 1920. “A Constant Error in Psychological Ratings,”Journal of Applied
Psychological 82 (5): p.665-674
Trompeter, G and A.Wright. 2010. “The World has Changed-Have Analytical Procedure
Practices?”Contemporary Accounting Research. 27 (2): p. 669-700
22
Appendix
Table 1. Matrix of Experiment
Data Compexity
Non-Complex
Knowledge Acquisition Method
Self-explanation (SE)
Explanatory
Without SE-EF
Feedback (EF)
Cell 1
Cell 2
Cell 3
Complex
Cell 4
Cell 5
Cell 6
Table 2 Result of One Way ANOVA test of the Effect of Demographic Characteristic on
Risk of Material Misstatement on Sales Account 2
Independent Variable
Sex
Age
Position
Working Period
Education
Public Accountant Training
Audit Training
Df
1
5
2
14
2
1
1
F-Statistic
0.154
1.122
0.121
0.651
0.658
0.006
0.084
Sig
0.697
0.367
0.886
0.803
0.430
0.938
0.773
Table 3 Testing of Experimental Error (Score of Accounting and Auditing Proficiency) on
Complexity Data and Risk of Material Misstatement on Sales Account 2
Independen Variable
Corrected Model
Intercept
Scope of information
Covariate Variable
Score of Accounting and Auditing Proficiency
Error
Total
Corrected Total
Sum of
Square
Type III
15.861a
18.910
15.820
Df
Mean of
Square
2
1
1
7.931
18.910
15.820
3.185
7.594
6.353
0.048
0.008
0.014**
0.585
154.385
1491.000
170.246
1
62
65
64
0.585
2.490
0.235
0.630
** Significant on 5%
23
F-Statistics
Sig
Table 4 Test of Hypothesis 1
Average (Standard of deviation)
Score of Halo Effect
Pretest
Posttest
(a)
(b)
Difference of
posttest – pretest
c = (b – a)
Cell 1, 2, 3
(N=32)
Non-complex
4.875
(1.539)
5.000
(1.437)
0.125
Cell 4, 5, 6
(N=35)
Levene Test
Complex
4.818
(1.776)
4.030
(1.686)
-0.788
F=0.048
Sign=0.827
T-test
t= 2.010
sign= 0.049**
Information:
a: Average (standard deviation) of material misstatement on sales account 1 (before obtaining data)
b: Average (standard deviation) of material misstatement on sales account 2 (after obtaining data)
c: Difference of Misstatement Risk of Sales Account 2 and Misstatement Risk of Sales Account 1
** significant at 0,05
Tabel 5 Testing of Hypothesis 2
Data
PANEL A
Cell 1 (N=10)
Cell 3
(control
group)
(N=11)
PANEL B
Cell 4
(N=11)
Cell 6
(control
group)
(N=12)
Mitigate
Strategy
Posttest
(a)
Pretest
(b)
Difference
of
Posttestpretest
(c)
Noncomplex
Selfexplanation
5.200
(1.619)
4.800
(1.989)
-0.400
(2.01)
Noncomplex
No
Mitigation
5.000
(1.265)
4.545
(1.635)
-0.455
(1.37)
Complex
Selfexplanation
3.727
(1.954)
4.090
(1.814)
0.363
(0.809)
Complex
Without
Mitigation
4.250
(1.712)
4.250
(1.712)
0.000
(1.651)
Difference
F levene test
(sign)
Levene test
F = 0.096
Sign= 0.760
t-test
t =0.073
Sign= 0.942
Levene test
F = 0.762
Sign= 0.393
t-test
t =0.660
Sign= 0.516
a: Average (Standard of Deviation) Risk of Material Misstatement Sales Account 2
b: Average (Standard of Deviation) Risk of Material Misstatement Sales Account 3
c: Difference of Average (Standard of Deviation) Risk of Material Misstatement Sales Account 3 - Average (Standard of Deviation) Risk of
Material Misstatement Sales Account 2
24
Table 6 Testing of Hypothesis 3
PANEL A
Cell 2
(N=11)
Cell 3
(control
group)
(N=11)
PANEL B
Cell 5
( N=10)
Cell 6
(control
group)
(N=12)
Scope
Information
Mitigating
Strategy
Posttest
(a)
Pretest
(b)
Difference
of Posttestpretest
(c)
Difference
F levene test
(sign)
Non-complex
Explanatory
Feedback
4.818
(1.537)
5.636
(0.809)
0.818
(1.078)
Levene test
F = 0.023
Sign= 0.881
Non-complex
Without
Mitigating
5.000
(1.265)
4.545
(1.634)
-0.454
(1.368)
Complex
Explanatory
Feedback
4.100
(1.449)
5.400
(1.955)
1.300
(0.948)
Complex
Without
Mitigation
4.250
(1.712)
4.250
(1.712)
0.00
(1.651)
****
25
t-test
t =2.442
Sign=
0.025**
Levene test
F = 0.309
Sign= 0.584
t-test
t =2.200
Sign=
0.040**
Halo Effect in Auditor Professional Judgment?
ABSTRACT
Previous research found that halo effect in audit context decrease quality of professional
judgment auditor. We propose that mitigating halo effect is necessary to improve accurate and
professional judgment of the auditor. We hypothesize that (1) Auditors with complex data would
determine the risk of material misstatement lower than auditors with non-complex data, (2) risk
of material misstatement for account that is determined by an auditor with self-explanation
higher than risk of material misstatement for account that is determined by an auditor without
self-explanation, (3) risk of material misstatement for account that is determined by an auditor
with explanatory feedback is higher than risk of material misstatement for account that is
determined by an auditor without explanatory feedback. Data from laboratory experiment using
various levels of auditors in Surabaya City Indonesia. The results show that halo effect occur
when auditor with complex data determine inaccurate professional judgment. The finding
suggests that shows that explanatory feedback as a strategy of mitigating halo effect, however we
do not find that self-explanation as a mitigating halo effect.
JEL: M42
KEYWORDS: Halo effect, complex data, non-complex data, self-explanation,
explanatory feedback, professional judgment
1
Introduction
AICPA guidance (2008) stated that assessing of client expectation is important phase in
analytical procedure. Analytical procedure is one of planning phase in understanding of client
business and industry. Holistic perspective in strategic assessment increases auditor attention to
client business risk and helps to increase the risk of misstatement (Bell, Peecher and Solomon,
1992). Research in psychology found that holistic evaluative in evaluating performance
associated with a holistic perspective potential to bias that known as the “ halo effect ”(Murphy
et al. 1993).
In audit context, the halo effect evidence support by O’Donnel and Schultz (2005) and
Grammling et al. (2010). The halo effect occurs when auditor assesses business risk in first step
of planning and then assesses risk of material misstatement that conform the value of business
risk (O’Donnel and Schultz 2005). These findings suggest that the holistic perspective auditors
develop during strategic assessment influences subsequent judgment. The other audit situation,
the halo effect occurs when knowledge about material weakness unrelated to the compensating
control encourages auditor to require a higher level of precision for a compensating control
(Grammling et al. 2010). Previous research investigate the predictor halo effect is holistic
perspective in evaluating an object. In audit context, we propose that the predictor of halo effect
is data complexity. Hogarth and Einhorn (1992) stated that the determinant of ambiguity is data
complexity, refers to the amount of information contained in each piece of data. If individual
weigh the judgment based on latest piece of data, known as recency effect. But, if the judgment
based on earlier piece of data is a primacy effect. Grick (2008) stated that halo effect is related
with primacy effect. So, this research proposes that the ambiguity of data caused halo effect in
analytical procedure.
2
Previous research (O’Donnel and Schultz 2005; Grammling et al. 2010) investigates the halo
effect in audit context but the mitigating strategy consider to research. Grammling et al. (2010)
stated that future research must considering mitigating strategy halo effect. This research
addresses the research question: how to mitigate halo effect in audit context? Cooper (1981b)
stated that there were several approaches to mitigate halo effect, for example, statistic control,
familiarity increasing by ratings, setting rating group, grouping irrelevant category, and rating
after observation and review. From those various methods, the method that is suitable for audit
context is training that is knowledge acquisition method. Training for beginner auditors is meant
to obtain complex analytical skill (Earley 2001). This is also in accordance with general standard
of first auditing stating that audit training should be done by a person or more who have adequate
skill and technical training as an auditor.
Beginner auditors learn procedural skills through personal experience or observation on other
more experienced auditors (Earley 2001). Research result from Earley (2001) extended Bonner
and Walker (1994) by conducting effectiveness testing of two knowledge acquisition methods,
which are self-explanation and explanatory feedback.
Self-explanation is a knowledge
acquisition method by giving individual an opportunity to explain to themselves the
rationalization answer that is chosen by them on some cases given. Explanatory feedback is
knowledge acquisition method by using an example and using it to solve a new case.
We experimentally examine auditor judgment in various levels (junior, senior, manager,
supervisor and partner) about the risk of material misstatement on sales account in noncomplexity and complexity data. The result suggests that the complexity data impact to auditor
professional judgment and the appropriate strategy to mitigate halo effect is explanatory
feedback. The next sections present literature review and hypotheses, followed by a description
3
of the experimental design. The results are then discussed, followed by conclusion and the
implication of research.
Literature Review
Thorndike (1920) defined halo effect as a marked tendency to think of the person in general as
rather good or rather inferior and to color the judgments of the (person’s specific performance
attributes) by this general feeling. Individual who conducts consideration for decision making
often has beginning assessment (anchor) on information that is then adjusted when getting new
information, that is called adjustment and anchoring heuristic (Hogarth, 1987). Auditor when
tends to weigh on last information is called recency effect, and when tends to weigh on the first
information is called primacy effect. Theory explaining primacy effect and recency effect is
known as belief adjustment theory that is developed by Hogarth and Einhorn (1992) who
considered ordering effect to examine interaction between duty characterization and information
process strategy. Belief adjustment model considers direction, power, and type of information
that by Bayes’ Theorem before, explained only decision making based on order and pattern of
information presentation. Halo effect, according to Grcic (2008), related to primacy effect
because the impressive anchor assessment was used as the assessment on additional new
information.
Individual does not revise their belief on additional new information because the
impressive anchor still attaches in their memory.
The Relation of Complexity Data and Risk of Material Misstatement
CPA firm develops a methodology for auditor to obtain, evaluate, and record evidence during
analytical procedure (Hirst and Koonce 1996; Trompeter and Wright 2006).
Holistic
information can be obtained from client or audit team leader after conducting understanding on
client’s business and given to senior auditor and junior auditor as a field assignment executor. If
4
auditor accepts holistic information that generally describes client’s condition that is convincing,
then in giving assessment on risk level of misstated accounting in analytical testing stage, it tends
to be in accordance with holistic information and the partner’s assessment.
This condition is supported that professional consideration done by experienced auditor would be
better than the one done by less experienced auditor (Knapp and Knapp, 2001). Executor auditor
as information provider tends to seek information that is consistent with his/her consideration,
meanwhile, partner as reviewer would tend to seek evidence that is inconsistent (Libby and
Trotman 1993). This can be explained by cognitive model of justification process which is
manipulation to seek evidence that supports other decisions or considerations (Peecher 1996).
Auditor combines objective justification in mental representation as quickly as they get, that
decision taken needs support (Gibbin dan Newton 1994). This condition occurs because in audit
methodology, decision making process is gradually, so the previous decision may affect the next
decision.
Phillips (1999) found that auditor who evaluating evidence related to low (high) - risk account
would be less (more) sensitive on detailed evidence of aggressive financial reports. Thus,
executor auditor when facing general information and obtaining strong impression on client’s
condition would tend to determine consideration that is consistent with the general assessment
when facing detailed transaction assignment.
Client’s condition that convinces and obtains low business risk value from partner as well as
good assessment of internal control system would cause auditor to determine low misstate risk.
Holistic information has formed mental representation on auditor about client who initially is
assessed positive. The limitation of auditor as an individual would tend to conduct assessment of
the next client generally similar to the beginning assessment. By representative bias, it can be
5
explained that general information is used as a comparison on the similarity of general
assessment with detailed attribute of financial reports in analytical assignment. When faced on
detailed assessment of accounting, auditor would use general information that is easily absorbed
in memory.
The level of material misstatement is considered high if auditor faces a client with high-risk
business. The determination of business risk based on the understanding on holistic information
of client that includes various aspects both from internal and external. Auditor facing holistic
information that is convincing on a client would tend to have high halo bias in determining the
risk of material misstatement in analytical testing. Client that is considered well by partner
would make mental representation in auditor that other evidences would in accordance with the
good beginning assessment. Auditor who conducts strategic assessment at high (low) level
would tend to determine the risk of material misstatement to be high (low) (O’Donnel dan
Schultz, 2005). Positive holistic on client’s condition in the form of high strategic assessment
could potentially cause high halo effect when facing account fluctuation inconsistency.
Hogarth and Einhorn (1992) introduced belief adjustment model that refers to the amount of
information contained in each piece of evidence. A greater number of the datum more
ambiguous unless a deeper level of cognitive processing eliminates the inaccurate hypotheses
(Luippold and Kida 2012). In managerial context, Booker et al. (2007) found that compared to
relative cost information increases designers’ focus on cost minimization. Specific cost
information were given the exact dollar cost of each design option. Dilla and Stone (1997)
suggest that individuals require more cognitive effort to acquire and use that are presented in
words.
6
Specific information that shows client’s condition in detail that there is the emergence of
fluctuation accounting inconsistence could potentially cause low halo effect.
Specific
information is pooled of data and can be presented in complex and non-complex data. Auditor
would be better in grasping the signal of account fluctuation inconsistency in specific
information (non-complex data). Therefore, implication of complex data in analytical assignment
which is the determination of high level risk of material misstatement. Based on the previous
argumentation and research, then, hypothesis one can be proposed as the following:
H1: Auditors with complex data would determine the risk of material misstatement lower than
auditors with non-complex data
The Relation of Self-explanation, Halo Effect, and Determination Risk of Material
Misstatement
Halo effect occurs when an auditor has an impression after facing holistic information, so it
affects mental representation in unconsciousness level. The effort of judgment bias mitigation is
done by emphasizing on alternative structure in evaluating information in decision making. The
effort is done because by general structure, it does not succeed (Balzer and Slusky 1992).
Research of consideration and decision making states that halo effect would affect sensitivity on
evidence that is contrary or not in accordance with holistic impression. Halo effect causes
individual to change the criteria used for evaluating and weighing decision information (Murphy
and Anhalt 1992). O’Donnel and Schultz (2005) gave empirical evidence in their experiment
when auditor conducting strategic assessment, then having halo effect, so they became
insensitive when facing account fluctuation that inconsistent in analytic testing.
Cooper (1981b) proposed training as one of methods to reduce halo effect. In accordance with
general standard of audit, to increase its professional consideration, CPA firm and Indonesian
7
Accountant Public Institute could hold training activities with various topics that in accordance
with audit assignment. Audit research shows that training could increase performance of audit
assignment including ratio analysis and real estate valuation (Bonner and Walker 1994; Earley
2001; 2003). Training context that can be used for auditors as the means of halo effect mitigation
is in analytical procedure because in this stage, halo effect could potentially emerges.
Determination of risk of material misstatement is done when an auditor conducts analytical
testing in planning stage. Auditors obtain information from both partner and convincing client,
so halo effect could potentially emerge. This bias affects auditor’s assessment when facing audit
evidence in detail. Therefore, the assignment of analytical procedure could be the focus of
training for halo effect mitigation.
Auditors who face complex data information have high potential of halo effect, so they tend to
be insensitive in determining the risk of material misstate.
Learning is a construction process
that makes a learner converses words and examples of instruction or that is presented in a text,
into skills that can be used to a problem solving (Chi et.al.1994). Learning is an acquisition and
knowledge organization through working memory that is a cognitive system to obtain and
manipulate information, intersection between long-term memory and information received
through senses, so it is able to be a significant component in the intellectual ability and the ability
problem solving (Brewster, 2011). By self-explanation knowledge acquisition method, auditors
with high halo effect could describe argumentation of risk level of misstatement given
previously. This description of process is a problem solving process, so the cognitive system
helps auditors to make a better professional consideration.
Based on argumentation and results of previous research, the third research hypothesis can be
proposed as the following:
8
H2: Risk of material misstatement with self-explanation is higher than risk of material
misstatement without self-explanation
The Relation of Explanatory Feedback Knowledge Acquisition Method, Halo Effect, and
Risk of Material Misstatement
Halo effect emerges because of holistic information received affects the professional
consideration when facing evidence in detail. Grammling et.al. (2010) emphasized the needs of
halo effect mitigation effort, so the quality of auditor professional consideration increased.
Kennedy (1993) explained the framework of unfamiliarity by using the method of training and
memory refreshing (internal data mitigation) and giving supporting instrument of decision
external data mitigation). Training by giving corrective feedback may mitigate halo effect
(Cooper 1981).
Balzer and Slusky (1992) showed that useful instrument to increase accuracy of decision making
is feedback. Lowe and Reckers (2000), Narsa (2010) used supporting instrument of foresight
decision to change auditor’s foresight perspective in order to approach hindsight perspective.
This is an effort to mitigate hindsight and foresight bias. In the effort of halo effect mitigation,
the use of decision supporting instrument that can be used is review result of audit manager on a
case faced by client.
Most of studies in psychology about learning were originally from working examples similar to
problems in the form of sentences on the textbook. Those examples are well-structured and each
had right answer objectively and solution procedure that can be accepted that was modeled in the
example.
Bonner and Walker (1994) showed that explanatory feedback and regulation
understanding showed that the result of cognitive psychology from learning that was wellstructured was also applied on audit domain that is ill structured. By explaining to auditors on
9
how to provide a right answer on a case helps them to obtain procedural knowledge. Training
may give good explanation before the problem, for example before working (similar to a case
given at the beginning of instruction) or after working (similar to a case after explanation
feedback).
Earley (2001) gave empirical evidence that an auditor who is given explanatory feedback would
have a better professional consideration compared to the one who does not get explanatory
feedback. Therefore, halo effect affects professional consideration, so when knowledge as the
form of explanatory feedback training method would increase professional consideration, as the
result, halo effect is mitigated.
Halo effect bias that is from individual limitation in processing information can be explained by
using heuristics representation. Individual tends to rely on information that has similarity level
with other information. To handle the bias, Bazerman (1994) explained the need of melting
process that changed status quo of individual decision making process in the form of emergence.
In explanatory feedback knowledge acquisition method, individuals would be faced to additional
information that can melt strong perception of memory attaching on the memory. Thus, halo
effect in risk determination of material misstatement can be reduced. Based on the argumentation
above, hypothesis 3 can be proposed as the following:
H3: Risk of material misstatement with explanatory feedback is higher than risk of material
misstatement without explanatory feedback.
10
METHODOLOGY
Experimental Design
The experimental design is pretest-posttest control group design. Shadish et al. (2002) explained
that the advantage of group control design with pretest and posttest is to prevent threat on
internal validity. An experiment is done by holding audit simulation in audit seminar. The
experiment in general can be described in research experiment matrix in table 1. The first factor
is data complexity consisting of two levels, non-complex and complex. The second factor is
method of knowledge acquisition that consists of self-explanation, explanatory feedback and
without self-explanation- explanatory feedback as a control group.
Experimental Subject
Experimental subject are auditors (junior, senior, manager and partner) in Surabaya City, the
second big city with large quantity of CPA firm in Indonesia. Randomization design is by
completely randomized design that by dividing two types of audit simulation modules to subjects
randomly. Therefore, each subject gets the same opportunity to be experimental group or control
group.
Operational Definition of Variable and Measurement
The complexity data has two levels, non-complex and complex data. The non-complex data was
adapted from Dilla and Stone (1997) and Booker, Drake and Heitger (2007). Participants with
non-complex data were presented the number of sales account information and words only,
without mentioning relative comparison. The complex data were presented in sales account
information in the form of number, words and relative comparison as Booker, Drake and Heitger
(2007).The words that explain the client situation in complex data contain explanation with
11
uncertainty condition. For example, the complex data stated that the increase of sales at 2012 is
17.76% relative with past year. The non-complex data stated that account of sales in 2012 has
increased. The risk of material misstatement on sales account in analytical test was measured by
the scores of 1 (very low) to 7 (very high). Participants determine material risk assessment in
three condition: (1) before get client data (risk of material misstatement on sales 1); (2) after get
client data (risk of material misstatement on sales 2), (3) after get mitigating strategy (risk of
material misstatement on sales 3).
RESULT AND DISCUSSION
Experiment
Sixty-five participants completed the study1. They were mainly staff-level auditor and partner
employed in Surabaya public accounting firms. The participants required to perform analytical
procedures on financial information with a seeded misstatement account. The participant got
simulation module consisting of module one for halo effect testing and complexity data and
module 2 for halo effect mitigation testing. The experimental design included two cases. Each
module was entitled and covered the same and had the same number of pages. The participants
consists of 38 junior auditor, 16 senior auditor, 3 manager, 6 supervisor and 2 partner.
Test of Demographic Characteristics on Risk of Material Misstatement on Sales Account 2
The risk of material misstatement of sales account 2 is determined after subject receiving
manipulation of holistic or non-complex data. ANOVA testing address to determine that risk of
material misstatement on sales account 2 is not affected by the difference of subject demographic
characteristic. The testing result by dependent variable was risk of material misstatement risk on
1
The participants invited to International Standard Auditing in Surabaya held on 2nd February 2013.
12
sales account 2 and independent variable that was demographic characteristic including sex, age,
position, working period, education, participation in public accountant training and audit training
show at table 2.
The result of ANOVA testing on the effect of subject demographic characteristic difference of
misstatement risk on sales account 2 showed that sex did not affect the risk of material
misstatement on sales account 2 (p=0.697). Age and position in CPA firm did not affect the risk
of material misstatement on sales account 2, and it was shown by probability value of age
(0.367) and position (0.886). Working period did not affect the risk misstatement on sales
account 2 that was shown by its probability value of 0.803. The last education, participation in
public accountant training and audit training also did not affect the risk misstatement on sales
account 2. The probability value of last education was 0.430, while public accountant training
was 0.938 and audit training was 0.773. Therefore, it could be concluded that risk of material
misstatement on sales account was not affected by the difference of subject demographic
characteristic. In experimental test, we must be careful with experimental error, i.e. accounting
and auditing capability of subjects. We test that difference of score of accounting and auditing
proficiency not affects risk of material misstatement on sales account 2. Testing of experimental
error (score of accounting and auditing proficiency) was done with ANCOVA (analysis of
covariate) with dependent variable is risk of material misstatement on sales account 2,
independent variable is scope of information and covariate variable is score of accounting and
auditing proficiency (Table 3).
13
Table 3 presents that score of accounting and auditing proficiency is not significant (0.630) and
scope of information as a manipulate variable is significant (0.014). The conclusion of this test is
risk of material misstatement on sales account 2 is only affected by scope of information and
does not affect the different score of accounting and auditing proficiency or demographic
characteristics.
Testing of Hypothesis 1
The result of hypothesis 1 testing is shown at table 4. The first hypothesis testing was done by
comparing the average of sales account 2 risk of material misstatement in the group with noncomplex data and the average risk of material misstatement on sales account 2 in the group with
complex data. Score pretest in cell with non-complex data is 4.875 and score posttest 4.875. The
distribution of non-complex data caused the increasing of sales account material misstatement
risk of 0.125. The opposite condition occurred in the group with complex data, with score
pretest 4.818 and 4.030, show the different of posttest-pretest the risk of material misstatement of
sales account -0.788. We compare the difference of risk of material misstatement on sales
account 2 and risk of material misstatement on sales account 1 between cell with complex data
and cell with non-complex data. The independent t-test shows that significance of that difference
is 0.049. This showed that subject with high halo effect and was given complex data, tend to
conform initial assessment.
The testing result showed that hypothesis 1 was supported by empirical evidence in this research.
Auditors who obtained non-complex data about the client would determine a lower risk of
material misstatement compared to the one who obtained non-specific data.
This finding
supported the occurrence of halo effect in the auditor receiving partner information about the
condition of client with complex or non-complex data. Higher halo effect and the complex data
14
from partner was still impressed by client’s well initial assessment, so the attached halo effect
would affect the risk assessment of material misstatement of sales account 2 to be similar to the
assessment of sales account 1 material misstatement risk. The opposite condition occurred on
auditor with high halo effect and obtaining information from partner with non-complex data.
Although it had positive impression in the initial assessment of the client, it would determine the
risk of material misstatement of sales account 2 higher than the risk of sales account 1 material
misstatement.
Testing Hypotesis 2
Hypothesis 2 Testing
Hypothesis 2 stated that self-explanation is a mitigation strategy to reduce halo effect. The
testing for pretest-posttest research design with control group was done by stages as the
following, first, determining the average difference of material misstatement risk of sales account
3 (posttest) with the average of material misstatement risk of sales account 2 (pretest), second,
comparing the difference of posttest-pretest in the group of non-complex data with the difference
of posttest-pretest in the group of complex data by independent t-test. The testing result of
independent t-test can be seen on table 5.
The testing result of independent t-test on the group of complex data showed that F value in
Levene Test was 0.096 with the significance value of 0.760 meaning that variances of both
groups were the same, while t-value was 0.073 with the significance probability of 0.942. This
finding showed that the average of material misstatement risk of sales account in cell 1 was not
significantly different with the average of material misstatement of cell 3 as control group.
15
The testing in the holistic group comparing the average difference of material misstatement risk
of sales account 3 to the average of material misstatement risk sales account 2 can be seen on
table 7. The testing result showed that F value in Levene test was 0.762 with the significance
value of 0.393 meaning that variances of both groups were the same, and t value was 0.660 with
significance probability of 0.516. This finding showed that there was no difference of material
misstatement risk in the group that was given self-explanation mitigation strategy with control
group that did not receive mitigation strategy. Therefore, it could be concluded that hypothesis 2
was not supported and self-explanation knowledge acquisition method did not get empirical
support as halo effect mitigation strategy.
Hypothesis 2 Discussion
Hypothesis 2 stated that self-explanation knowledge acquisition method would reduce halo effect
in determining material misstatement risk was not supported by empirical data. It means, selfexplanation did not affect significantly on the determination of accurate material misstatement
risk when auditor faced complex data that triggered high halo effect. The result of this research
did not succeed proposing self-explanation knowledge acquisition method as a mitigation
strategy to reduce the effect of halo effect in the determination of material misstatement.
The essence of self-explanation was to give auditor opportunity to realize the consideration that
was done and able to absorb information given by partner with complex data. Furthermore,
auditor was able to determine the risk of material misstatement more accurately compared to the
misstatement risk determined previously.
In this research, auditor with high halo effect was expected to be able to conduct self-explanation
and to revise professional consideration that had been given. This research gave the fact that
auditor who conducted self-explanation did not succeed to revise the professional consideration
16
that had been done. This was probably because the auditor was not able to realize their own
problem without other’s help.
Other findings in the group with self-explanation mitigation strategy was that it did not give right
rational argumentation with the case, in fact, some people answered only with one sentence
without trying to argue logically. The list of questions consisting 8 questions to memorize
condition of the client delivered in holistic or non-complex data of partner’s information given
before subject gave an argumentation. All questions were answered by the subject with the
average answer of 6.6, meaning that subjects were able to memorize well the information given.
However, some subjects when asked to give an argumentation, answered with only one or two
sentences.
Hypothesis 3 Testing
The existence of explanatory feedback knowledge acquisition as the mitigation strategy to reduce
the effect of halo effect in determining material misstatement risk was detected by determining
the difference of material misstatement risk average of sales account 3 (posttest) and the average
of material misstatement sales account 2 (pretest), and next, comparing the difference of posttestpretest risk average of non-complex data group. Theoretically, the group with complex data or
non-complex data that had high halo effect would change its assessment when receiving
explanatory feedback compared to the control group that did not receive explanatory feedback.
Hypothesis 3 testing was done by comparing the difference of material misstatement risk average
of sales account 3 with the average of material misstatement risk of sales account 2 for each
information scope group with independent t-test.
17
The testing showed that there was the significant difference between cell 2 that received
information scope and explanatory feedback mitigation strategy and cell 3 that received noncomplex data and without mitigation strategy. Statistically, it could be seen that the value of
t=2.422 with significance probability of 0.025 for the group with non-complex data.
The result testing in the group of complex data and receiving explanatory feedback mitigation
strategy showed that there was a significant difference between cell 5 (explanatory feedback) and
cell 6 (without mitigation). The result of t-test showed that t value=2.200 with p=0.04. The
testing showed that explanatory feedback knowledge acquisition method could be used as a halo
effect mitigation strategy, thus, the professional consideration taken was more accurate.
Hypothesis 3 Discussion
Explanatory feedback knowledge acquisition method would reduce halo effect in determining
material misstatement as hypothesis 3 was supported by empirical data. This finding was in line
with the previous research that explanatory feedback could increase the accuracy of decision
making (Balzer and Slusky, 1992; Earley, 2001; Bonner and Walker, 1994).
Halo effect occurred when the auditor had impression on the client’s condition that was
convincing and obtained partner’s information with complex data affecting mental representation
in unconsciousness level. The effect of halo effect occurring was the decreasing of sensitivity on
the evidence that was contrary or did not fit the first impression. In this research, the auditor was
asked to understand the business and industry of the client by watching a video and reading a
booklet of client’s company profile. The detection of halo effect was done by giving questions
to the auditor to give an initial assessment on minimarket management, distribution system and
financial performance of the client before receiving financial and non-financial information in
detail. The finding shows the majority of auditor had high halo effect.
18
The next stage was to ask the auditor understanding information from the partner about sales and
selling cost account and to conduct analytical procedure that compared quantitative and
qualitative information as well as compared the account balance in 2011 and 2012.
The
identification of the halo effect attachment when obtaining partner’s information so it affected
the accuracy of decision making could be seen from the material misstatement risk that was
determined in the group that received complex data with the risk of material misstatement on the
group that received non-complex data.
This research showed that the risk of material
misstatement determined in the group of complex data was lower than the risk of material
misstatement in the group of non-complex data.
The result of good initial assessment on the client was used as the basic to make decision when
conducting the next assessment (explaining representativeness heuristic). This could be reduced
by giving explanatory feedback in the form of manager’s review result containing the case
preview of client especially on the account fluctuation of sales and selling cost. The finding of
this result that was explanatory feedback could increase the accuracy of professional
consideration, so it reduced halo effect, strengthened the statement of Cooper (1981) that training
with corrective feedback could mitigate halo effect.
Conclusion
Initial judgment in evaluate convincing client profile will influence the subsequent judgment.
The result of this research found that halo effect occur in analytical procedure when auditor
determine professional judgment with complex data. Auditor with non-complex data will assess
risk of material misstatement lower than auditor with complex data.
19
Self-explanation knowledge acquisition method as halo effect mitigation strategy could not be
supported in this research. This finding showed that method of self-explanation with written
argumentation cannot change the first opinion and halo effect cannot mitigation. Cognitive
process from auditor’s self when conducting learning on a problem, could not be separated from
self-desire to learn and the willingness to analyze the problem.
The result of this research gave empirical evidence that explanatory feedback in the form of
manager’s review result was a knowledge acquisition method that could be halo effect mitigation
strategy on auditor. Halo effect that could be mitigated was marked by the increasing of
professional consideration accuracy in determining material misstatement risk. This finding was
consistent with (2001, 2003) that explanatory feedback could increase auditor’s professional
consideration.
This result also gave contribution that halo effect could be mitigated by
explanatory feedback.
This study has three limitations. First, about the experimental case materials just use positive
halo effect. Further research can extend with negative halo effect and determine risk of material
misstatement. Second, the context of analytical procedure is in initial hypothesis. Further
research is expected to improve the experiment materials in another audit test, i.e. in substantive
test or in analytical procedure in final evaluation phase.
Third, this research focus in an
individual judgment, other research may be developed in group decision making.
References
American Institute of Certified Public Accountants (AICPA). 2012. “AICPA Professional
Standards.” New York: AICPA
Bazerman, M.H. 1994. Judgment in Managerial Decision Making. Singapore: John Wiley and
Sons.
Balzer, W. and M. Sulsky. 1992. “Halo and Performance Appraisal Research: A Critical
Examination, “ Journal of Applied Psychology, p. 975-985
20
Booker, D. M., A. R. Drake, and D. L. Heitger. 2007. New Product Development: How Cost
Information Precision Affects Designer Focus and Behavior in a Multiple Objective
Setting. Behavioral Research in Accounting. 19: 19-41
Bonner, S. E., & P. L.Walker. 1994. The Effects of Instruction and Experience on the
Acquisition of Auditing Knowledge. The Accounting Review. 1: 157-178
Brewster, B. E. 2011. How a System Perspective Improves Knowledge Acquisition and
Performance in Analytical Procedures. The Accounting Review. 86 (3): 915-943
Cooper, W. H. 1981. “Ubiquitous Halo,”Psychological Bulletin.vol.90. no.2: p. 218-244
Chi, M.T.H., N. de Leeuw, M.H. Chiu, and C. Lavancher. 1994. Eliciting Self- Explanation
Improves Understanding. Cognitive Science. 18: 439-477
Dilla, W. N., and D.N.Stone. 1997. “Representations as Decision Aids: The Asymmetric Effects
of Words and Numbers on Auditors’ Inherent Risk Judgments,”Decision Sciences 28
(3): p.709-743
Earley, C. E. 2001. Expertise Acquisition in Auditing: Training Novice Auditors to Recognize
Cue Relationships in Real Estate Valuation. The Accounting Review.1: 81-97
______, C. E. 2003. A Note on Self Explanation as a Training Tools for Novice Auditors: The
Effects of Outcome Feedback Timing and Level of Reasoning on Performance.
Behavioral Research in Accounting. 15: 111-124
Gibbins, M, and J.D Newton. 1994. “An Empirical Exploration of Complex Accountability in
Public Accounting,”Journal of Accounting Research. (Autumn): p.165-186
Grammling, A., O'Donnel, E., and S.Vandervalde. 2010. “Audit Partner Evaluation of
Compensating Controls: A Focus on Design Effectiveness and Extent of Auditor
Testing,”Behavioral Research in Accounting. Vol. 29: p. 175-187
Hogarth, R.M. 1987. Judgment and Choice.2nd Edition. Singapore: John Wiley & Sons
Hogarth, R.M., and H.J. Einhorn. 1992. “Order Effects in Belief Updating: The BeliefAdjustment Model,” Cognitive Psychology, Vol. 24: p.278-288
Hirst, D., and L. Koonce. 1996. “Audit Analytical Procedures: A Field Investigation,”
Contemporary Accounting Research: 457-486
Libby, R and K.T.Trotman. 1993. “Audit Review as a Control for Biased Recall of Evidence in
Decision Making,”Accounting, Organizations and Society.18: p.559-574
Kennedy, J. 1993. Debiasing Audit Judgment with Accountability: A Framework and
Experimental Result. Journal of Accounting Research. 31: 481-491
Knapp, C., and Knapp, M. 2001. “The Effects of Experience and Explicit Fraud Risk Assesment
in Detecting Fraud with Analytical Procedures,” Accounting, Organization and
Society.vol.26: p.25-37
Luippold, B. L., and T.E. Kida. 2012. “The Impact of Initial Information Ambiguity on the
Accuracy of Analytical Review Judgment,” Auditing: A Journal of Practice & Theory. 31
(2): p.113-129
21
Lowe, D. J., and P. M. J. Reckers. 2000. The Use of Foresight Decision Aids in Auditor’s
Judgment. Behavioral Research in Accounting. 12: 97-118
Murphy, K., Jako, R.and R.L. Anhalt, 1993. “Nature and Consequences of Halo Error: A Critical
Analysis. Journal of Applied Psychology,” vol 2: p. 218-225
Murphy, K.R., and R.L Anhalt. 1992. Is Halo Error a Property of the Rater, the Ratees, or the
Specific Behavior Observed? Journal of Applied Psychology. 77: 494-500
Narsa, I, M. 2010. Alat Bantu Keputusan Foresight dan Strategi Mitigasi Bias Hindsight untuk
Mempersempit Kesenjangan di antara Perspektif Auditor dan Evaluator. Disertasi.
Universitas Gadjah Mada, Yogyakarta
O' Donnel, E. and Schultz, J. J. 2005. “The Halo Effect in Business Risk Audits: Can Strategic
Risk Assesment Bias Auditor Judment about Accounting Details?” The Accounting
Review, Vol 80, No 3: p. 921-939
Peecher, M.E. 1996. “The Influence of Auditor’s Justification Processes on Their Decisions: A
Cognitive Model and Experimental Evidence,”Journal of Accounting Research. 34(1):
p.125 – 140
Philips, F. 1999. “Auditor Attention to and Judgment of Aggressive Financial Reporting,”
Journal of Accounting Research. Vol 37. No.1. Spring
Shadish, R.W; Cook, T and Campbell, D.T. 2002. Experimental and Quasi-Experimental Design
for Generalized Causal Inference. Houghton Mifflin Company
Thorndike, E. L. 1920. “A Constant Error in Psychological Ratings,”Journal of Applied
Psychological 82 (5): p.665-674
Trompeter, G and A.Wright. 2010. “The World has Changed-Have Analytical Procedure
Practices?”Contemporary Accounting Research. 27 (2): p. 669-700
22
Appendix
Table 1. Matrix of Experiment
Data Compexity
Non-Complex
Knowledge Acquisition Method
Self-explanation (SE)
Explanatory
Without SE-EF
Feedback (EF)
Cell 1
Cell 2
Cell 3
Complex
Cell 4
Cell 5
Cell 6
Table 2 Result of One Way ANOVA test of the Effect of Demographic Characteristic on
Risk of Material Misstatement on Sales Account 2
Independent Variable
Sex
Age
Position
Working Period
Education
Public Accountant Training
Audit Training
Df
1
5
2
14
2
1
1
F-Statistic
0.154
1.122
0.121
0.651
0.658
0.006
0.084
Sig
0.697
0.367
0.886
0.803
0.430
0.938
0.773
Table 3 Testing of Experimental Error (Score of Accounting and Auditing Proficiency) on
Complexity Data and Risk of Material Misstatement on Sales Account 2
Independen Variable
Corrected Model
Intercept
Scope of information
Covariate Variable
Score of Accounting and Auditing Proficiency
Error
Total
Corrected Total
Sum of
Square
Type III
15.861a
18.910
15.820
Df
Mean of
Square
2
1
1
7.931
18.910
15.820
3.185
7.594
6.353
0.048
0.008
0.014**
0.585
154.385
1491.000
170.246
1
62
65
64
0.585
2.490
0.235
0.630
** Significant on 5%
23
F-Statistics
Sig
Table 4 Test of Hypothesis 1
Average (Standard of deviation)
Score of Halo Effect
Pretest
Posttest
(a)
(b)
Difference of
posttest – pretest
c = (b – a)
Cell 1, 2, 3
(N=32)
Non-complex
4.875
(1.539)
5.000
(1.437)
0.125
Cell 4, 5, 6
(N=35)
Levene Test
Complex
4.818
(1.776)
4.030
(1.686)
-0.788
F=0.048
Sign=0.827
T-test
t= 2.010
sign= 0.049**
Information:
a: Average (standard deviation) of material misstatement on sales account 1 (before obtaining data)
b: Average (standard deviation) of material misstatement on sales account 2 (after obtaining data)
c: Difference of Misstatement Risk of Sales Account 2 and Misstatement Risk of Sales Account 1
** significant at 0,05
Tabel 5 Testing of Hypothesis 2
Data
PANEL A
Cell 1 (N=10)
Cell 3
(control
group)
(N=11)
PANEL B
Cell 4
(N=11)
Cell 6
(control
group)
(N=12)
Mitigate
Strategy
Posttest
(a)
Pretest
(b)
Difference
of
Posttestpretest
(c)
Noncomplex
Selfexplanation
5.200
(1.619)
4.800
(1.989)
-0.400
(2.01)
Noncomplex
No
Mitigation
5.000
(1.265)
4.545
(1.635)
-0.455
(1.37)
Complex
Selfexplanation
3.727
(1.954)
4.090
(1.814)
0.363
(0.809)
Complex
Without
Mitigation
4.250
(1.712)
4.250
(1.712)
0.000
(1.651)
Difference
F levene test
(sign)
Levene test
F = 0.096
Sign= 0.760
t-test
t =0.073
Sign= 0.942
Levene test
F = 0.762
Sign= 0.393
t-test
t =0.660
Sign= 0.516
a: Average (Standard of Deviation) Risk of Material Misstatement Sales Account 2
b: Average (Standard of Deviation) Risk of Material Misstatement Sales Account 3
c: Difference of Average (Standard of Deviation) Risk of Material Misstatement Sales Account 3 - Average (Standard of Deviation) Risk of
Material Misstatement Sales Account 2
24
Table 6 Testing of Hypothesis 3
PANEL A
Cell 2
(N=11)
Cell 3
(control
group)
(N=11)
PANEL B
Cell 5
( N=10)
Cell 6
(control
group)
(N=12)
Scope
Information
Mitigating
Strategy
Posttest
(a)
Pretest
(b)
Difference
of Posttestpretest
(c)
Difference
F levene test
(sign)
Non-complex
Explanatory
Feedback
4.818
(1.537)
5.636
(0.809)
0.818
(1.078)
Levene test
F = 0.023
Sign= 0.881
Non-complex
Without
Mitigating
5.000
(1.265)
4.545
(1.634)
-0.454
(1.368)
Complex
Explanatory
Feedback
4.100
(1.449)
5.400
(1.955)
1.300
(0.948)
Complex
Without
Mitigation
4.250
(1.712)
4.250
(1.712)
0.00
(1.651)
****
25
t-test
t =2.442
Sign=
0.025**
Levene test
F = 0.309
Sign= 0.584
t-test
t =2.200
Sign=
0.040**